Now that Google I/O 2019 is completely over, let's take it back to 2018, when the company announced a couple of neat new features for Google Photos: Color Pop and Colorize. Color Pop, which lets users easily apply a black-and-white filter to a picture's background to give the subject a nice "pop," was deployed just days after I/O 2018. Colorize, meanwhile, went under the radar for a whole year. Today, the company is finally updating us on this front and is teasing a beta program to try the tool out.
The self-explanatory Colorize feature was introduced last year as a trick to bring old black-and-white photos to life. It was touted as one of the improvements made possible with TensorFlow machine learning.
Of course, a 20-second demo doesn't tell the whole story on the research and development of Colorize, but what doesn't help is the fact that literally no news was made about it between then and now.
What got Google finally talking? On Twitter, Photos product lead David Lieb reacted to a call-out on the company's radio silence on Colorize and said that the feature was still a work in progress.
2/ Here’s a photo of my 104yo grandmother on her wedding day, colorized with Google Photos on my phone. (You can see we have some work to do; my grandfather didn’t wear pink pants to his wedding!) pic.twitter.com/Ni8v0Bz3vg
— David Lieb (@dflieb) May 6, 2019
When color is everything you're trying to achieve in transforming a picture, the bar for accuracy is quite high and the job's hard enough for humans to perform. AI developers might find some help in using a new tool called TCAV — highlighted at I/O this year — to analyze how they can better train their neural networks' "vision."
Lieb followed that post with another saying that his team hopes to open up a Beta to test the feature, but there's no timeline.